11 Replies Latest reply on Nov 21, 2011 11:12 AM by severud

    NPM 10.2 Beta 3 - Thoughts / Issues / Observations

    jspanitz

      Figured I'd start a running thread and just keep adding to it as we dig deeper into Beta 3

      • So far, as I indicated in another thread - our Junipers aren't supported :(
      • Discovery keeps timing out and we are only trying to discover 5 class c subnets.  We extended the timeout to 4 hours and it still fails, well before the 4 hours.
      • Added the beta NPM server as a managed node.  Perhaps this is not new to the beta but never noticed it before, the Node Details Hardware first indicated it was Physical and the Node Details Description / Location / Contact was blank.  Of course this did correct itself after subsequent polls, but that info was all set on the server from the start.
      • According to the blog post here - Orion NPM vNext Sign Up for Beta & – Tell us what you think on the feature Multi-CPU Polling - we should be seeing "As of the next version, you will be able to report, alert and visualize these types of machines by CPU. So we do not plan to take away any of the existing resources, reports etc. You will still have the gauge that does the average and the Min, Max and Average CPU Utilization chart. What we will be adding is two new resources, as illustrated below in the first image, that will break out each CPU’s usage over time in a chart, but also the current value in a table format."  We are probably just missing it, but we don't see those resources anywhere.

      That's it for now.  Going to attempt to get discovery working so we can test the Improved Topology.....

        • Re: NPM 10.2 Beta 3 - Thoughts / Issues / Observations
          jspanitz

          Ooops.  Posted to the wrong forum.  Can a moderator move this to the NPM Beta forum.  Thx!

          • Re: NPM 10.2 Beta 3 - Thoughts / Issues / Observations
            Halef

            So far, as I indicated in another thread - our Junipers aren't supported

             Do you mean CPU and Memory polling, subinterfaces, or both?

            Discovery keeps timing out and we are only trying to discover 5 class c subnets.  We extended the timeout to 4 hours and it still fails, well before the 4 hours

            Please let us know if you manage to solve this. If not, please collect diagnostics and open a support case so we can look at the issue.

            Added the beta NPM server as a managed node.  Perhaps this is not new to the beta but never noticed it before, the Node Details Hardware first indicated it was Physical and the Node Details Description / Location / Contact was blank.  Of course this did correct itself after subsequent polls,

            This behavior is by design. The polling scheduler now distributes polls evenly in time. This information is collected during rediscovery poll, which can happen anytime in 0 - 30 minutes after the node is added. You can use "Rediscover" button on the node details to make it run earlier.

            • Re: NPM 10.2 Beta 3 - Thoughts / Issues / Observations
              smartd
              Beta 3 failed to install.  Stopped during install, with a "service OrionModuleEngine failed to start.  Verify that you have sufficient privledges to start services.  Of course, I'm Local Admin on the machine.  Case # 279552.  I should have waited for GA.  This stuff happens every time I test beta code.
              • Re: NPM 10.2 Beta 3 - Thoughts / Issues / Observations
                smartd
                How exactly does the Layer 3 Topology work?  What resource is needed to show topology for a node?  I can't see a change from the old topology feature.
                  • Re: NPM 10.2 Beta 3 - Thoughts / Issues / Observations
                    mkarak

                    Hi,

                    There are no new topology resources introduced in 10.2. We changed the way how we collect information about topology (we are using separate L2 and L3 polling jobs that are polling topology data from devices with each rediscovery interval and background process that is calculating connections from this data).

                    Could you please check if devices that should be connected using L3 have "Topology: Layer 3" poller assigned? You can perform "List Resources" on particular device to see assigned pollers.

                    All discovered connections are displayed in "NPM Network Topology" resource. If you missing some L3 connections, please collect diagnostics and open support ticket so we can investigate. Please do not forget to include information about connections that are not identified by Orion.

                    One more thing: Are you running 10.2 Beta3 or 10.2 RC3? Beta3 was release 2 moths ago to limited set of customers and was not officially supported (regarding installation on production environment). We are currently in RC phase (RC3 released last week). There were several changes since Beta3 (including topology improvements). If you actually running 10.2 RC3, please post your findings, questions, requests, etc. into our RC forum, which is located here:

                    http://thwack.solarwinds.com/forums/59/release-candidates/179/orion-network-performance-monitor-release-candidate/

                    Thanks, Milan

                    • Re: NPM 10.2 Beta 3 - Thoughts / Issues / Observations
                      Karlo.Zatylny

                      Hi Dan,

                      While we are gathering Layer 3 information from devices (ARP tables), we are not currently giving any actual Layer 3 topology edges.  The edges in NPM are still driven from the layer 2 connections reported by the switches (Bridge MIB).  When we can't find a connection using the MAC address reported by the Bridge MIB, we use the layer 3 entries to hopefully find an edge that way.  10.2 has better topology from the standpoint that the new methodology finds more edges than the previous one and does not require you to have all your topology giving nodes in a discovery together.  The latter point making discoveries faster as well.

                      We are interested in the accuracy of the edges so that we can make improvements to the new algorithm, but the changes may be that if you were getting reasonably good edges in the previous versions, you might not see much of an improvement.  Where we were good before, we seem to be at least as good still.

                      Thanks